Background Recent candidate and genome-wide association studies have identified variants altering susceptibility to breast cancer.
Objective To establish the relevance of these variants to breast cancer risk in familial breast cancer cases both with and without BRCA1 or BRCA2 (BRCA1/2) mutations.
Methods A cohort of unrelated individuals with breast cancer due to the presence of either BRCA1 (121) or BRCA2 mutations (109) and individuals with familial breast cancer not due to BRCA1/2 mutations (722) were genotyped using Taqman SNP Genotyping Assays. Allele frequencies were compared with an ethnically and gender-matched group (436).
Results A synonymous variant (Ser51) in TOX3 (previously TNRC9) was associated with an increased risk of breast cancer (OR=1.82, p<0.001) in BRCA2 mutation carriers. The associations for FGFR2 (OR=1.20, p=0.046), TOX3 (OR=1.5, p<0.001), MAP3K1 (OR=1.26 p=0.03), CASP8 (OR=0.73 p=0.02) and the chromosome 8-associated SNP (OR=1.31, p=0.004) were replicated in individuals without BRCA1/2 mutations. In addition, homozygote carriers of MAP3K1 variants were shown to have a significantly lower Manchester Score (mean 13.8–17.6, p=0.003), whereas individuals carrying one or two copies of the FGFR2 variant had a higher Manchester Score (mean 17.5–17.9, p=0.01).
Conclusions This study confirms that susceptibility variants in FGFR2, TOX3 and MAP3K1 and on chromosome 8q are all associated with increased risk of cancer in individuals with a family history of breast cancer, whereas CASP8 is protective in this context. The level of risk is dependent on the strength of the family history and the presence of a BRCA1/2 mutation and contributes to the understanding of the use of these variants in clinical risk prediction.
- Cancer: breast
- familial breast cancer
- genome wide association study
- Manchester Score
Statistics from Altmetric.com
- Cancer: breast
- familial breast cancer
- genome wide association study
- Manchester Score
Population-based epidemiological studies have shown that mutations in BRCA1 and BRCA2 genes account for only 16–20% of familial breast cancer cases.1 Despite the identification of other breast cancer susceptibility genes of high to moderate penetrance including TP53, PTEN, ATM, BRIP1, CHEK2 and PALB2, the genetic basis of 80% of familial cases remain unexplained.1 Evidence from twin studies and the clustering of breast cancer in families suggested that multiple susceptibility genes, conferring small genotypic risks, collectively accounted for the majority of risk in the population.2 Numerous candidate gene association studies have been conducted to establish the contribution of genetic variants to breast cancer susceptibility.3 However, the results of these candidate gene studies have often been conflicting, and robust, replicable susceptibility variants have not emerged.4 An exception is the common missense variant, aspartic acid to histidine (Asp302His) in CASP8, which has been associated with a moderate reduction in breast cancer risk and delayed onset of cancer in BRCA1 and BRCA2 carriers.5 6 Recent advances in genotyping technology and statistical analysis have facilitated the conduct of large-scale and genome-wide association studies (GWASs).7 GWAS are able to analyse hundreds of thousands of single-nucleotide polymorphisms (SNPs) in a single assay from both women with breast cancer and healthy controls, and multiple comparisons of allele frequencies made. SNPs are chosen either on the basis of their potential functional relevance—for example, coding changes within genes (cSNPs)—or through the selection of common tagged SNPs, which represent common haplotypes, to ensure even coverage across the entire genome. Further, GWASs have the advantage of being relatively unbiased in that they consider variants independent of any a priori evidence supporting a relationship with breast cancer pathogenesis. In 2007, Easton et al8 undertook a three-stage GWAS to identify common variants associated with breast cancer susceptibility. Four loci contained plausible causative genes (FGFR2, TOX3, MAP3K1 and LSP1), with no gene candidate at the associated 8q24 locus. Each locus conferred small increased risks ranging up to 1.23-fold in individuals carrying one copy of the variant (heterozygotes) and 1.63-fold in those with two copies (homozygotes). An independent GWAS also reported the association with an intronic variant in FGFR2 in a sporadic postmenopausal breast cancer population.9 A subsequent case-only study defined phenotypic correlations for these variants, such that individuals carrying a MAP3K1 variant were less likely to be lymph node positive at diagnosis, individuals with the TOX3 variant were more likely to be diagnosed at a younger age, and individuals with a variant in FGFR2 had a stronger family history of breast cancer and/or ovarian cancer.10 A further analysis of the original GWAS cohort revealed that FGFR2 and the 8q24 locus were more strongly associated with oestrogen receptor (ER)-positive disease.11 Other breast cancer GWASs have provided evidence for association with chromosomal regions on 6q22.33, 2q35, 16q12 and 5p12.12–14
A further study has investigated the role of variants in three of the associated genes (FGFR2, TNRC9/TOX3 and MAP3K1) in increasing breast cancer risk in over 10 000 BRCA1 and BRCA2 mutation carriers.15 Variants in FGFR2 and MAP3K1 were each associated with increased breast cancer risk in BRCA2 mutation carriers, but not in BRCA1 carriers, whereas the TOX3 variant was associated with increased breast cancer risk in both BRCA1 and BRCA2 mutation carriers. Importantly, the loci interacted multiplicatively on breast cancer risk in BRCA2 mutation carriers.
We aimed to establish the relevance of the variants in FGFR2, LSP1, CASP8, TOX3 and MAP3K1 plus the chromosome 8q variant to the risk of breast cancer in individuals with a strong family history of breast cancer either with BRCA1 and BRCA2 mutations or who were known not to carry mutations in these two genes.
Females with a strong family history fulfilling the NICE criteria for BRCA1 and BRCA2 screening (>20% risk of mutation) mutation analysis between 1990 and 2007 at the Regional Genetics Service, St Mary's Hospital, Manchester were recruited.16 A series of lower-risk samples were tested on a research basis as previously reported.17 The entire coding sequences of BRCA1 and BRCA2 were screened by DNA sequence analysis and by multiplex ligation-dependent probe amplification for deletions and duplications. The familial breast cancer cohort comprised two groups: 230 unrelated white British women (median age=42.3 years, range=20–81) with BRCA1 or BRCA2 mutations and 722 unrelated white British women (median age=46 years, range=18–78) without BRCA1 or BRCA2 mutations. It is important to note that 63 of the 722 familial cases were submitted to the GWAS of Easton et al8, but none of the BRCA1 or BRCA2 mutation-positive cases were represented in the analysis by Antoniou et al.15 The control group comprised up to 436 unrelated white British women (median age=39 years, range=16–82) without cancer and with no immediate family history of cancer. All samples were provided with informed consent for clinical genetic testing and anonymised for research studies as approved by the North Manchester research ethics committee (08/H1006/77).
Genotyping was carried out using pre-designed rs3817198 (LSP1), rs889312 (MAP3K1), rs3803662 (Ser51, TOX3), rs2981582 (FGFR2) and rs13281615 (Chr.8-variant) and custom-designed rs1045485 (Asp302His, CASP8) Taqman SNP Genotyping Assays (Applied Biosystems, Warrington, UK). Samples that failed or provided equivocal results with Taqman Assays were genotyped by DNA sequencing analysis. Of the Taqman genotyped samples, ∼5% were selected for sequencing analysis and found to be 100% concordant. The primers were designed on the basis of the variant sequences in GenBank (accession numbers NM_000465, NM_001013253, NM_005921, NM_001080430, NM_000141, NT_008046, NM_001228), and provided by Invitrogen Ltd (Paisley, UK) (for primer sequences see supplementary table 1). Minimum genotyping success was 98% and 87% for familial breast cancer cases and controls, respectively (the latter was due to use of some poor quality, older DNA samples).
The study was pragmatic, designed on the basis of the number of samples available. For each of the SNPs under consideration, we had power ranging from 8.5% to 50.6% in the BRCA1/2 mutation-positive group and 11.1% to 74.8% in the BRCA1/2 mutation-negative group, with 95% confidence in each to detect the ORs reported in the original GWAS8 or the CASP8 association study5 (table 1).
Consistency with Hardy–Weinberg equilibrium (HWE) was established by χ2 tests. Genotype associations were assessed using logistic regression on the genotypes and number of alleles, and expressed as ORs with 95% CIs. BRCA1, BRCA2 and non-BRCA cases were considered separately, and each group compared with controls, and BRCA1 compared with BRCA2. Associations between genotypes in subgroups with high and low Manchester Score (using a cut-off value of a Manchester Score of 16 to define the two groups) are considered in a similar manner, and allelic ORs computed comparing low and high scorers. Formal tests of the association between genotype and Manchester Score were compared using one-way analysis of variance of the score between the three genotypic groups. Corrections for multiple comparisons were not made, as this was a replication study.
Control genotypes for all the SNPs considered in this study were consistent with HWE (p>0.05) except for FGFR2 (p=0.002). Therefore 10% of all the samples were genotyped for the FGFR2 variant by DNA sequence analysis; 100% concordance between Taqman and sequence results was noted, excluding genotyping error as an explanation for the discordance. The control cohort was genotyped for a number of other disease susceptibility variants, including CARD15, IL23R, ATG16L1 and BARD1 (data not shown); allele frequencies were consistent with published Caucasian control data and no deviations from HWE were noted. Therefore the control population was considered to be representative of a normal healthy female population without a family history of cancer.
BRCA1- and BRCA2-positive cohort
The Ser51 synonymous variant in TOX3 was associated with increased breast cancer risk in patients with BRCA2 mutations (p<0.001, OR=1.82, 95% CI 1.29 to 2.57), but not in BRCA1 carriers (p=0.31, table 2). BRCA2 carriers with one TOX3 variant risk alleles had a twofold increased risk of breast cancer (p=0.0014, OR=2.08, 95% CI 1.32 to 3.28), whereas BRCA2 carriers with two TOX3 variant risk alleles had a nearly threefold increased risk of breast cancer (p=0.02, OR=2.73, 95% CI 1.18 to 6.29). No positive associations were demonstrated between the other susceptibility variants and breast cancer in the BRCA1 and BRCA2 mutation-positive cohort (table 2).
Familial breast cancer cohort without BRCA1 or BRCA2 mutations
Approximately 700 individuals with breast cancer and a strong family history, but without BRCA1 or BRCA2 mutations, were genotyped for the susceptibility variants. Allelic associations between increased cancer risk and the FGFR2 (p=0.046), TOX3 (p<0.001), MAP3K1 (p=0.03) and chromosome 8 variants (p=0.004) were replicated (table 3).
In addition, the previously reported protective effect of the CASP8 variant was replicated (p=0.02). No positive association between the variant in LSP1 and breast cancer was established. Similar to the BRCA2 mutation-positive group, the strongest association was for the synonymous variant in TOX3. Two copies of the TOX3 variant conferred a 2.5-fold increased risk (p=0.002, OR=2.39, 95% CI 1.39 to 4.09).
Stratification of the BRCA1/2 mutation-negative cohort
To establish if the susceptibility variants altered the degree of familial risk, the BRCA1/2 mutation-negative cohort was stratified into two groups by using the Manchester Score as a surrogate for the strength of the familiality (table 4). The Manchester Score estimates the likelihood of an individual carrying a BRCA1/2 mutation based on the number of cases of breast and ovarian cancer in a family and the age of female breast cancer disease diagnosis.17–19 Family history of male breast cancer, prostate and pancreatic cancer also contributes to the prediction model. A cut-off score of 16 was used; this is equivalent to a 10% chance of identifying a BRCA1/2 mutation combined.17–19 This score would be achieved by having four individuals in a family with breast cancer between 50 and 59 years of age or two aged 30 and 39 years.17
The variant in MAP3K1 was associated with a higher risk of breast cancer in individuals with a lower Manchester Score (ie, less familial disease). Individuals homozygous for the MAP3K1 variant have a significantly lower Manchester Score than wild-type or heterozygote individuals (p=0.003). Conversely, individuals heterozygous or homozygous for the minor allele of the FGFR2 variant had a higher Manchester Score (p=0.010). The chromosome 8 variant was more strongly associated with breast cancer in the group with lower Manchester Scores, but this difference did not meet significance. There were no significant differences in Manchester Scores between individuals heterozygous or homozygous for variants in TOX3, CASP8 or LSP1 (table 4).
Recent GWASs and candidate gene studies have identified a number of genetic variants, which confer a modest increased risk of breast cancer.5 8 We have conducted a validation study in a cohort of individuals with a strong family history of breast cancer, who had undergone analysis for the presence of BRCA1 and BRCA2 mutations. We were able to confirm the association between the variant in TOX3 and breast cancer in individuals with BRCA2 mutations previously reported by Antoniou et al.15 Our study was underpowered to detect the other reported positive associations between variants in FGFR2 and MAP3K1 in carriers of BRCA2 mutations and the variant in TOX3 in individuals with BRCA1 mutations. However, the CIs for the ORs for these variants overlapped with those reported by Antoniou et al.15 In contrast with the previous report, which detected that the strongest association was with the FGFR2 variant, in our study the association between TOX3 and breast cancer was strongest in BRCA2 mutation carriers. The association between the variant in TOX3 and breast cancer in individuals with BRCA2 mutations supports increasing evidence of important molecular differences between BRCA1- and BRCA2-related breast tumours,20 including the finding that BRCA2-related breast cancers are more likely to be ER positive than those with BRCA1 mutations.21 This finding is also consistent with a prospective study that showed that the variants in both TOX3 and FGFR2 are associated with ER-positive disease.11
In the cohort of patients with familial breast cancer without BRCA1 and BRCA2 mutations, we replicated the associations between variants in FGFR2, TOX3 and MAP3K1 and the variant on chromosome 8 and increased breast cancer risk.8 15 We also established a significant association between the protective missense variant in CASP8 and breast cancer in the mutation-negative familial cohort. This supports the protective effect of this variant in familial breast cancer.6 Similar to the finding in the BRCA2 mutation-positive group, the strongest association in our study was between the TOX3 variant and increased risk of breast cancer. Individuals heterozygous for the minor allele had a 1.5-fold increased risk, whereas homozygotes had a 2.4-fold increased cancer risk. Despite the relatively small numbers of cases and controls considered in our study, these results demonstrate the power of using a highly selected population with a positive family history to validate the results of GWASs. The study design applied the reciprocal approach of Easton et al, who used a highly selected population with familial disease for the initial screen to detect associations in their GWAS.8 The studies reported so far confirm the relevance of breast cancer susceptibility variants in patients both with and without a family history of breast cancer and with and without BRCA1 and BRCA2 mutations.8 15 22–24
We employed the Manchester Score to establish the likelihood of an individual carrying a BRCA1 or BRCA2 mutation.17–19 The associations between the variant in MAP3K1 was stronger in the groups with the lower Manchester Score, whereas the FGFR2 variant was more strongly associated with increased risk in the group with a higher Manchester Score. These results indicate that breast cancer susceptibility variants will have different contributions to familial disease either in altering the level of risk or prognosis. The different degrees of association for FGFR2 and MAP3K1 mirror the findings of Rebbeck et al24 in a cohort of unselected individuals, which demonstrated that FGFR2 was associated with breast cancer in European–American and African–American populations, whereas MAP3K1 was only associated in the latter. This type of information will become increasingly important as clinical algorithms are generated to incorporate SNP data to determine specific risks for individuals with a family history of breast cancer.25 26 It is possible that families with high Manchester Scores still have as yet unidentified moderate risk alleles, which account for part of the familiality, with less necessity for aggregation of lower-risk alleles in a polygenic model. The families with lower scores could be accounted for solely by aggregation of lower-risk alleles. Our study confirms the relevance of the genetic susceptibility variants in individuals with a strong family history of breast cancer and establishes that the risks conferred are similar to those in individuals from case–control series of sporadic breast cancer.
Common variants in a number of genes have been associated with an increased risk of breast cancer.
We demonstrate that variants in FGFR2, TOX3, MAP3K1 and on chromosome 8q are all associated with an increased risk of cancer in individuals with a family history of breast cancer.
A variant in CASP8 is protective in individuals with a family history of breast cancer.
Highly selected familial cohorts are powerful for replicating the results of genome-wide association study results.
The Department of Medical Genetics is supported by the NIHR Manchester Biomedical Research Centre.
Web Only Data 47/2/126
An additional table is published only at http://jmg.bmj.com/content/vol47/issue2
Funding This study was funded by Genesis UK, who had no role in study design; in the collection, analysis and interpretation of data; in the writing of the manuscript; nor in the decision to submit the manuscript for publication.
Competing interests None.
Ethics approval This study was conducted with the approval of the North Manchester Research Ethics Committee 08/H1006/77.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.